Voice classifiers distinguish COVID groups in a small held-out sample
Source check: Source checked. Not independent scientific confirmation.
Finding: A study trained voice classifiers in 210 participants and tested their combined predictions in 30 additional participants. The system classified 24 of 30 correctly, including all 10 positive cases. The selected symptomatic, Italian-speaking sample does not establish asymptomatic screening performance or population-level accuracy.
Technology: Radial-basis support vector machines on acoustic features
Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- SVMs learned acoustic distinctions between clinically defined groups from extracted voice features; voting combined task-specific predictions.
- Test and result
- Three cohorts of 70 participants supplied voice recordings for feature selection and binary SVM training. Separately recruited test participants comprised 10 positive,10 recovered and 10 healthy individuals. Table 5 reports 24/30 correct,10/10 positive,8/10 healthy and 6/10 recovered. Majority voting combined three tasks and three binary comparisons.
- Limitation / still unknown
- Training feature selection/ranking precedes the described ten-fold evaluation, so internal cross-validation is not clearly nested; repeated recordings also need participant-level separation. The 30-person test is a small balanced within-study sample, not independent replication. Table 5 includes one uncertain recovered participant, despite an abstract statement that all errors were classified positive. Most positive/recovered patients were symptomatic, and language/ethnicity restrictions constrain generalization. Long-term vocal effects and surveillance benefit were not demonstrated.
Evidence sources and annotation context
Evidence context (descriptions follow the cited sources; test type is a catalog classification):
Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8616736/
Source checked · editorial details
The source page was checked against this entry. Independent confirmation is not assessed by this label.
Source type: paper. Paper: linked separately.
Review date: 21 Nov 2026. Review scope: Primary full-text bibliographic dates, computational classification Methods, held-out Results and explicit cohort/reference limitations..
No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.
Editorial assessment: Impact 1/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.